方法对比

	Scripture Stylometry	Computational Stemma Reconstruction
领域	Religious Studies	Religious Studies
方法族	Machine learning	Machine learning
起源年份	2002	2009
提出者	John Burrows (Delta); applied to scripture by Faigenbaum-Golovin et al. and others	Adapted from biological phylogenetics (Howe, Robinson, O'Hara); benchmarked by Roos & Heikkilä
类型	Distance-based stylometric model over function-word frequencies	Algorithmic tree-inference pipeline for reconstructing manuscript genealogies
开创性文献	Burrows, J. (2002). 'Delta': a Measure of Stylistic Difference and a Guide to Likely Authorship. Literary and Linguistic Computing, 17(3), 267-287.	Roos, T., & Heikkilä, T. (2009). Evaluating methods for computer-assisted stemmatology using artificial benchmark data sets. Literary and Linguistic Computing, 24(4), 417-433.
别名	Stylometric Analysis of Sacred Texts, Computational Stylistics of Scripture, Burrows's Delta for Scripture, Quantitative Stylistics of Religious Texts	Phylogenetic Stemmatology, Computer-Assisted Stemmatology, Algorithmic Stemma Building, Cladistic Textual Criticism
相关	4	4
摘要	Scripture stylometry measures the writing style of sacred texts quantitatively, chiefly through the frequencies of the most common words, in order to compare passages, detect authorial layers, and test traditional claims about who wrote what. Its workhorse is John Burrows's Delta, introduced in 2002, which represents each text as a profile of standardized function-word frequencies and measures the stylistic distance between texts as the average difference between those profiles. Because function words such as articles, prepositions, and particles are used unconsciously and at rates that vary little with subject matter, they form a stable stylistic fingerprint. Recent work, such as the 2025 word-frequency study of the Hebrew Bible by Faigenbaum-Golovin and colleagues, shows how these techniques distinguish scribal corpora and corroborate or challenge the layers identified by traditional source criticism.	Computational stemma reconstruction borrows the mathematics of biological phylogenetics to rebuild the family tree of a manuscript tradition automatically from coded variant readings. Each surviving witness is treated as a taxon and each place of textual variation as a character with discrete states, exactly as a biologist treats species and the genes that vary among them. Tree-inference algorithms then search for the genealogy that best explains the observed pattern of variants, typically the tree requiring the fewest reading changes (maximum parsimony) or the most probable tree under an evolutionary model. Teemu Roos and Tuomas Heikkilä's 2009 study established how to evaluate these methods rigorously, building artificial manuscript traditions with a known true stemma and measuring how accurately each algorithm recovered it. The result is a scalable, reproducible complement to the hand-built Lachmannian stemma.